Severe injury to brain may be expected to produce a nonlinear response in evoked potentials (EP). Timely detection of such injury-related changes may help in diagnosing patient's status in critical care and surgical monitoring situations. Coherence function is used to measures the degree of linear association between the EP signals recorded during normal and abnormal experimental conditions. An adaptive coherence estimation algorithm is developed to estimate time-varying model parameters responsive to the brain injury. A linearity index is devised so as to obtain a single measure of the brain's departure from linearity. Analyses of somatosensory EP show a very sharp drop in the magnitude coherence estimates during hypoxic injury and a corresponding rapid decline in the linearity index at the very early stages of the hypoxic injury.